The following content has been provided by the University of Erlangen-Nürnberg.

So good morning everybody. Today we will close the section on distortion correction.

Before we do that, let's look at the contents of diagnostic medical image processing.

What did we consider so far? What are the topics we have seen so far?

And what are the core messages that you should take with you out of the discussions?

So first of all, we talked a little bit about different modalities.

And that's a very superficial introduction.

We just learned a little bit about the modalities that are used in medical imaging.

But we did not take the time here to discuss all the physics in detail.

For us, these are cameras and they generate images.

And then we look at the images and try to get the best out of these images that are acquired by various modalities.

Just to name a few, standard and mostly known modalities are X-ray, CT, MR, or NMR.

It's sometimes called in the literature nuclear magnetic resonance imaging.

The N was canceled by the time because nuclear sounds very dangerous.

Or MRI, magnetic resonance imaging is the term that is mostly used. Then we have seen that ultrasound, endoscopy, PET,

positron emission tomography, and SPECT, that's single photon computed tomography, and many more.

So that's just a rough overview. And by the time we will learn to assign the corresponding modality to the images we see,

you will learn more and more how to read these images.

And then I told you that the whole lecture for the winter term is divided up into three subsequent chapters.

One chapter is on preprocessing.

So we look basically at the problem, what can we do to the images to improve the image quality for later processing or for diagnosis.

Then we will talk about what can I do if I have multiple images of one and the same objects.

And we will talk about gaining higher dimensional information by looking at CT reconstruction methods.

This will be taught by this semester. This will be taught by Andreas Meyer.

So he will show up here in about two or three weeks and then he will teach you the secrets of image reconstruction.

And right after Christmas, we will think about the problem, what do I do if I have multiple images from multiple modalities,

and I want to combine all the image information.

And this chapter is called Image Fusion. We will learn here a lot of algorithms, how to deal with it.

And in the chapter on preprocessing, we are currently in one less important part.

We deal with image intensifiers and artifacts that are implied by image intensifiers.

And if you see this double E, sorry, double I, then you immediately have in mind this picture.

This is a bucket and this is the detector here, the whole thing that transforms basically the incoming X-ray particles, photons into an electronic signal, into an image, right?

And the old technology of image intensifiers is basically carrying here a vacuum tube with an electron optics.

So the photons, they generate electrons, the electrons are accelerated in an electron optics,

and then the energy of the electrons is transformed into a visible image that is captured by a standard CCD camera.

And the problem is that within this vacuum tube, that's about 30 centimeters or so,

we have these electrons moving in the earth's magnetic field, so these are deviated,

and these deviations cause image distortions and they have to be corrected.

And we looked into that and I explained to you that using a calibration pattern, we can capture an image of the distorted calibration pattern.

We know how the calibration pattern should look in the image if no distortion appears.

And based on these correspondences of points, we get undistorted point, distorted point.

We estimate a mapping between the distorted and the undistorted image.

I also explained to you in which direction this estimation or this mapping should be done.

We talked about parametric mappings to parameterize the mapping, then we estimated the parameters using standard linear algebra.

And this is a chapter here we squeezed in right after the introduction.

That was a chapter on SVD to explain to you the important properties or the basic properties of SVD.

If you haven't heard this, this was hopefully a very efficient introduction.

If you heard about this in previous lectures, that was hopefully a concise refresher course on it.

So we know how to estimate these parameters and we talked about very important things that you should extrapolate or you should apply in the future also to many other problems that go well beyond image distortion and might not be related to image distortion at all.

These techniques are very general.